Gabe Giro

RankRush is an answer-engine-optimization platform that helps businesses track and improve their discoverability in AI assistants. It spans 3 modules: AI visibility tracking, Reddit community engagement (Buzz), and social ads (Pulse).

As a fractional CTO and full-stack engineer, I led the migration off Lovable onto self-managed infrastructure, our own Supabase (PostgreSQL, Deno Edge Functions, RLS), VPS hosting, an Astro marketing site, and a Python CLI for tooling. Day to day I owned strategy, drafting and posting while my co-founder Nikola owned the research pipeline and visibility scoring.

Developed skills and plugins for the team through Claude Enterprise.

Yahoo Fantasy Sports is a fantasy sports app delivered in three months. I was in charge of updating features in the Daily Fantasy part of the app while also updating the code from Java to Kotlin and various architectural patterns to MVVM with RxJava. Also, I built guides for the refactoring for other developers to follow.
• Built my tasks according to product directions.
• Used Agile and Kanban with daily scrum meetings.
• Helped update different developer guides and Jira ticket templates and helped optimize processes inside the team.

I developed two premium content-streaming Android applications and their libraries, and my most significant contribution was making the UI code more robust. Until then, the UI had received more than two bug fixes per week. After that, we changed the UI only if the specifications changed.

Technologies: Android, TDD, Java, Mockito, PowerMockito

HungerStation is a food delivery app. The company had 50k+ team members while I was in constant communication with around 10 people, most common being the Product Owner, Designer, iOS engineer and 2 backend engineers.

I was in charge of developing a self-contained Flutter feature for robotic delivery from scratch with cross-functional team. The goal was to have minimum impact on the rest of the code while following the Dart and Flutter best practices and team's cookbook.

The delivery was successful with all the required features done. I achieved this with TDD which gave me much more confidence in handing over the code to maintainers afterwards. And I got the biggest compliment an engineer can get from a fellow engineer - that I have a very neatly organized and functioning code.

Most of my work on this project consisted of updating the UI and implementing Android Wear API for an urgent release of the podcast application. This was my first Kanban project, made in Android Studio, and hosted on GitHub.

Technologies: Android, Android Wear

I was in charge of implementing a screen reader accessibility feature from scratch. Working there was one of my best work experiences.

I successfully finished adding the accessibility feature to every screen.

I participated in the five-people team that Toptal Projects provided for the client, including one senior and junior delivery manager, project manager, and UX/UI designer. I communicated daily with two product owners, one DevOps, and two technical support persons, leading another Android senior engineer.

• I acted as a go-to tech person for this project, performing as an Android architect/lead engineer.
• Took responsibility for research and development and took critical project risks. Researched the market and watched by requirements of the client.
• Set up epics and stories for the MVP in collaboration with the team
• Built a roadmap for delivery.
• Wrote the setup guide for the app to work correctly in the Kiosk mode on the smartwatch.
• Built native calls functionality with custom UI (not VoIP).
• Helped hire a great addition to our Toptal Projects team—a senior Android engineer—who complemented my skills perfectly.

Zufa is an AI parental tool that builds children's critical thinking rather than just surveilling them. It shipped as a Chrome extension monitoring YouTube, with a next phase targeting Safari, iOS, Android, Mac, and Windows on a production backend.

As a half-time fractional CTO, I shepherded the product toward production readiness:
• Owned the Safari Web Extension TestFlight publishing pipeline (cloud-managed distribution cert and App Store Connect API).
• Built cost-transparency reporting on the Google Gemini usage so investors could see unit economics.
• Set platform strategy across the cross-surface mix while keeping the MV3 extension architecture clean.

The hardest call was sequencing the platform expansion without over-investing before the back end was ready.

Kotlin and Java are two powerful general-purpose languages popular for Android and beyond. I wrote a blog post where I discussed their top features and differences, then focused on how to smoothly transition between the two.

I worked with Plume on an enterprise app called Plume Workpass. I was part of an Android team of 10 and more than 100 people working on this app.

This was my first project, which involved clean architecture and a Kotlin multiplatform. I contributed to almost every part of the app—from the data layer through the domain and presentation to the UI. One very useful concept I learned was the mappers between each layer in the clean architecture.

A local-first IDE for my own Obsidian vault, built solo in about a week with Vite, React, and Express.

• Rendered the entire vault with wikilinks and Mermaid diagrams, plus purpose-built views: a task Kanban/Eisenhower matrix, a housing dashboard, and a relationship-graph canvas.
• Wired Claude streaming chat with live vault-context injection, push-to-talk voice input via whisper.cpp, and TTS playback for replies.
• Made files hot-reload through chokidar and server-sent events, so editing a note in Obsidian updated the IDE instantly.
• Shipped the POC (v0.1.0) direct-to-main as a single author, after deliberately scrapping an earlier Next.js cloud-PWA prototype in favor of a faster localhost-first architecture.

An app for mastering the art of drawing and sketching.

• Hired to speed up the app development.
• Built the camera and AR functionality from scratch. OpenGL was previously used, and I replaced it with ArCore.
• Placed images in the real world and handled their movement, scale, and rotation.
• The code had been previously built with the help of ChatGPT, but without proper architecture in mind. That presented many interesting challenges, especially given the client's very limited budget. So, I had to focus on navigating the existing code and adding the functionality without refactoring.
• Wrote a guide to refactor the code through a few steps for the client.
• After the main functionality was implemented, the client extended the contract to launch the app in beta and production.
• The most challenging task was handling focus with ArCore, as I needed to use fixed focus and let the user refocus. We almost moved away from ARCore until I had an opportunity to talk to another dev, who pointed me in the right direction of the SharedCamera feature.

I worked on this mother-targeted app that ranked number one for downloads in its country. I co-developed the Android version from scratch and was mainly in charge of the UI. Delivering the required customization made this a fun project.

Technologies: Android, Analytics

I was in charge of bug fixing and implementing new features.
It was a one-month engagement, but in that time, I fixed all the bugs that came my way and helped finish one beta feature.

Jumprope was acquired by Linkedin in 2021.

TAKH Tutor AR is an AR/VR app that educates patients about their healthcare products.

I was the lead Android engineer using Minds and Assembly in a five-person team. I also worked closely with an in-house iOS developer and a 3D designer to match the experience. Another Android developer and I were in charge of finishing the Android app within tight deadlines.

Project Highlights:
• Developed my first commercial app with Kotlin and ARCore library.
• Built custom controllers that utilize only a drag gesture to successfully interact with an object in 3D space.
• Animated multiple VR/AR objects that moved with a click. I increased my knowledge of the technology tremendously throughout the project to make this possible.

I led a 7-engineer mobile team at Journey Rewards, a loyalty platform for travelers, as we transitioned development in-house from an outsourced vendor. I drove the technical strategy and accelerated the delivery process by implementing build checks, automated and Claude-assisted pull request reviews, and per-feature ownership while continuing to develop Android, iOS, and Kotlin Multiplatform Mobile (KMM) features.

This app is a personalized weight-loss program and I worked as a lead Android engineer in a team of five for two months.

The biggest challenge was to make changes in a legacy code and an architecture that didn't follow MVP principles as intended.

I worked as the sole senior Android engineer building an MVP of the app from scratch to publish, and I was in charge of publishing updates. This was one of two enterprise apps I developed for FocustApps.

Technologies: Android, Google Maps, REST, Git, HockeyApp, libGDX (3D), Firebase, Google Cloud Messaging

I was the sole senior Android engineer building the MVP of this app from scratch and publishing it along with updates. This was one of the two enterprise apps I developed for FocustApps.

Technologies: Android, Google Maps, REST, Git, HockeyApp, libGDX (3D), Firebase, Google Cloud Messaging

This was a fintech company with aspirations to become a private hedge fund.
I managed a team of four highly driven people, including one trader, one sales representative, one quant algorithm engineer, and an engineer in charge of automatizing processes.
• We developed an EA (expert advisor) with a hedging algorithm in MQL4. 90% of direction guesses are successfully closed in profit.
• I was in charge of organizing meetings, standups, work sessions, PR reviews, occasional EA feature development, accounting, and much more.
• Kanban organized in Trello with daily scrum standups.
• This was a side project for all of us.
• The biggest challenge was to get everyone on the same page and charge for the same goals.

I built an autonomous product factory that turns ideas into shipped software with little to no human intervention. I combined a Hetzner VPS orchestrator (agent-army) with a product factory toolchain that sandboxes Docker builds, gates spend, and routes Slack messages into automatically deployed code changes. The platform operates in two modes: a weekly product pipeline and managed surfaces that maintain existing repositories through sandboxed editing, building, deployment, and code pushes.

A scientific research app for a skin disease called HVS.
I made the app from scratch with a Clean version of MVVM (Model View ViewModel) architecture.
Although I did regularly publish the app to Beta testers via Firebase Releases, I didn't stay on board to publish it live

A tool for finding IG models to attend events like charity, awards, and photoshoots.

I tried building my model for gender recognition through GCP’s VisionAI and VertexAI, again based on two Kaggle datasets and a CelebA dataset. After unsuccessful attempts, as that was not good enough data to train on, I went to test five Hugging Face models. The best, with 90% accuracy, was https://huggingface.co/NTQAI/pedestrian_gender_recognition.

I built a data pipeline with SOAX Proxy, created five fake IG accounts to avoid being blocked, and wrote Python scripts for different stages of ETL.

At one point, I found 10,000 new unique profiles in 30 minutes. My biggest win was optimizing the SQL UPDATE command with a combination of STRUCT, MERGE, and UNNEST, where the GCP updates went from increasing minutes (1 to 30 mins) and a minimum of 10 GB per query, down to a few seconds and 10s of MB.

While building IGRecruiter, I was going through Udemy’s course “Python for Machine Learning & Data Science Masterclass” by Jose Portilla. I had two mentors to help me navigate through the challenges and learn faster. I had weekly meetings with each.

Letro is a service like email, but end-to-end encrypted, spam-free, phishing-free, and works without the Internet, too—all thanks to Awala, a delay-tolerant overlay network.

I was in charge of setting up the project from scratch in a clean variation of an MVVM architecture without the domain layer.

Recall is a Rust-based task manager by Clemens Winter; I built the Android port.

• Preserved the entire Rust domain layer untouched and exposed it to Android through a JNI bridge, then built a native Jetpack Compose UI on top, so the Android app and the desktop TUI shared one source of truth.
• Implemented full peer-to-peer sync between desktop and Android using Iroh (QUIC) with ALPN version negotiation, so devices reconnected across app restarts without re-pairing.
• Shipped 28 PRs upstream into the Rust repository plus direct-to-main commits on the Android app, working across Rust (recall-core, recall-jni, ratatui), Kotlin, Coroutines, and SQLite event sourcing.

An automated bot that applies to LinkedIn jobs via Easy Apply based on your preferences.

It started as a fork of an existing bot with few fixes to the outdated functionality (due to LinkedIn's HTML changes).

The primary motivation was to make it easy for tech-savvy individuals with basic knowledge of Python and Docker. The secondary motivation was to add answers to questions as part of the free functionality compared to the original repo.

I also developed a user-friendly, closed-source solution called JobHunter.

WarmLink is a platform designed to promote referral marketing through LinkedIn connections.

I was responsible for full-stack development of the front end, back end, and the Chrome extension. I developed a data pipeline that included a Chrome extension to scrape LinkedIn contacts, which were then pushed to the back end on GCP. The database provided users with their LinkedIn connections on the front end, allowing them to send referral marketing messages through the platform.

An eCommerce webshop that sells food and beverages worldwide.

The client had two years of sales data in a CSV from WooCommerce orders and one year of tracking with Google Analytics, which presented interesting challenges.

There were two goals:
• Figure out where to target the next campaign to maximize profits.
• Detect performance issues and UI/UX bottlenecks.

Here is the Git repository for cleaning anonymized data: https://github.com/GabeGiro/witrina.

Some of the tasks during data cleaning were:
• All transactions before 12th Aug 2022 from Croatia were in HRK; converted to EUR to match the rest of the purchase history.
• All transactions after 31st Dec 2022 were in EUR.
• Order shipping was sometimes empty. I asked the client, and there were three reasons: (1) the user used a coupon, (2) a user from Croatia ordered more than 52 EUR of products, and (3) users from the rest of the world ordered more than 290 EUR of products.

Some of the insights shared with the client:
• Order shipping was not properly applied to some of the transactions.
• A few comparisons were added to discuss the next potential target country, city, customer profile, and product.

A full-stack project for job application automation that leverages EasyApplyJobsBot (a bot that searches and applies to EasyApply jobs on LinkedIn). I was responsible for full-stack development and decision-making. I built the back end with Python, the front end with Next.js, and a PostgreSQL database, and connected everything with Docker Compose.

The app is still unreleased, but I can showcase it on a call.

Idyllic was in a pre-investment stage, and they needed a well-documented code review. In one week, I wrote more than 20 pages in a Google Doc file with a heavy emphasis on security. In addition to the in-depth security review, I also gave code improvement suggestions, composed bug reports, and showed design improvement according to Android design guidelines.

I delivered a speech titled "From Android Beginner to a 5-star Freelancer." This was my first professional talk, and I enjoyed speaking and sharing my knowledge about freelance work. After the speech, I obtained one business deal and a few developers approached me for a job, which was great.

The link attached is an article written about the talk (in Croatian).

I co-founded the TaxiCentral startup. Our goal was to make a super easy-to-use app for people who want to order a taxi in Zagreb. Basically, it was like Uber for taxi companies.

I was responsible for the entire development, rollout, and everything in between—working on the architecture, documentation, partial design, testing, publishing, and promotion. This was my first published application.

Technologies: Android

I developed the Android part while my college peers implemented the web crawling for this project. We won first place and received other perks like free office space for a year. The link attached is our pitch for the project (in Croatian).

BookBottles helps nightclubs easily manage reservations, table inventory, guest lists, client profiles, and transactions. I implemented the communication between the client and our developer on this project.

I built an Android app from scratch with a Parse-based back end for a pub-crawling brand for this project. I also set up everything so the founders could easily add events to the Parse back end.

Technologies: Android, Parse

The client's goal was to publish an Android pub-crawling app to the Google Play Store. They chose me to develop it after I reviewed the two previous developers' code.

I implemented the full development of the Android and Parse back end, completing the weekly tasks I had brainstormed with the client. I also migrated the app from Eclipse to Android Studio and the back end from DreamFactory to Parse.

Technologies: Android, DreamFactory, Parse, Facebook integration

The client's goal was to complete a partially (70%) completed app. They had previously employed four freelance developers who each, in turn, did some work before moving on. The code was in terrible condition and the client had a very tight schedule.

I worked on the UI changes, database, and REST API without improving the code because of the tight schedule and limited budget. The client delivered a fully developed app on time.

Technologies: Android, SQLite

I was in charge of building a POC (proof of concept) app in Flutter. This was my first professional project in Flutter, and I was amazed at how fast you can develop features and easily connect with the back end. Unfortunately, before I could finish the app, the budget for the project ran out.

Occasionally, working on this app helps keep my priorities straight.
I'm learning Flutter and Firestore with it. It's a simple app that can have multiple projects and tasks inside each. On the home screen, you see only the highest-priority tasks from each project.

I built a YNAB-style household finance operating system with a calm, luxury-inspired UI. I implemented budget buckets (Ours, John, and Jane) with income-share splits, an automated buffer, and FIRE tracking, along with multi-currency support for EUR and RSD and a local-first ML transaction categorizer. I developed both the web and native SwiftUI (local-first) applications on a PostgreSQL back end embedded in the app image and imported approximately 5,630 real YNAB transactions.

Developed alongside two senior Android engineers. I was in charge of bug-fixing and implementing new features.

Although I was hired for short-term help, I learned a lot about Android Jetpack, Dagger 2, and Coroutines.